Great Circle Associates List-Managers
(February 2003)
 

Indexed By Date: [Previous] [Next] Indexed By Thread: [Previous] [Next]

Subject: Re: false positives
From: James Lick <jlick @ drivel . com>
Date: Wed, 26 Feb 2003 20:21:37 +0800 (CST)
To: list-managers @ greatcircle . com
In-reply-to: <20030226113300 . 21046 . qmail @ houston . wolf . com>
References: <tneff @ grassyhill . net> <130798312 . 1046214147 @ [192 . 168 . 254 . 79]> <3E5BF111 . 13482 . A908AAE @ localhost> <1399 . 1046242169 @ kanga . nu> <20030226113300 . 21046 . qmail @ houston . wolf . com>

On Wed, 26 Feb 2003, Angel Rivera wrote:
> To that we have added SpamAssassin in tagging mode.  The few false positives
> that it catches are simply tagged and can be whitelisted.

SpamAssassin 2.50 which was released last week also supports Bayesian
filtering.  Bayesian filters can be trained on a per user basis what is
and isn't spam according to the individual user.  After being taught, it
is a fairly accurate counterweight to the standard SpamAssassin rules.  If
you find a message that has incorrectly been tagged as spam or not spam,
just use sa-learn, and the weightings will be adjusted in the future.

Using previous versions of SpamAssassin, I had few false positives (some
real Chinese mails, etc.), the new version no longer flags those as spam
thanks to the Bayesian filters.  I'm getting much fewer false negatives to
boot.

PS- Anyone managed to wedge SpamAssassin into any list manager?  It would
be really nice if MailMan just chucked out spams straight away rather than
asking me to slog through rejecting them each morning.

---- James Lick ---- jlick @
 drivel .
 com .
 tw ---- http://drivel.com.tw/ ----


Follow-Ups:
References:
Indexed By Date Previous: Re: false positives - junkmail
From: "Angel Rivera" <angel @ wolf . com>
Next: Re: false positives
From: James Lick <jlick @ drivel . com>
Indexed By Thread Previous: Re: false positives
From: "Angel Rivera" <angel @ wolf . com>
Next: Re: false positives
From: James Lick <jlick @ drivel . com>

Google
 
Search Internet Search www.greatcircle.com